Review of "Assessing the consistency of satellite derived upper tropospheric humidity measurements" by Shi et al.

An assessment of four satellite upper tropospheric humidity (UTH) data sets are made with a focus on interannual variability and the effect of El Nino Southern Oscillation (ENSO) on spatial anomalies. Although the spatial anomalies associated with ENSO variability are well known, the strength of this article is in comparing 4 contrasting datasets. In general, users of these or similar datasets will probably wish to know the best dataset for their application so any statements on the relative quality of the datasets and any issues such as inhomogeneities or artificial drifts will add value to the paper. In particular, are there any spurious jumps or trends such as HIRS before and after 2000 or jumps in the CMSAF record? It would also be useful to comment on whether there are systematic biases for any particular meteorological regimes (e.g. convective, anvil, descent, etc).

Specific comments:

Abstract: some more quantitative statements would be useful in assessing the magnitude of differences.

L45 - application for model evaluation and processes understanding studies could be highlighted

L131 - presumably non-linearities in the UTH calculation affect the computation of UMIAMI UTH from gridded data and this will affect the absolute magnitude but probably not the anomalies

Fig.1 - it is expected that HIRS UTH will be lower than microwave estimates since they sample systematically drier, clear-sky scenes. Can the lower UMIAMI values compared to FIDUCEO and CMSAF be explained by the method, in which case why is UMIAMI used as the baseline since it is not computed using swath data?

L147 and throughout - more quantification of the difference between datasets would be helpful

L153 - please provide quantification of "good agreement"

L155 - how close is "close to each other" in the context of their use?

L157 (and L256) - is the smaller variability in HIRS linked to the fact it is not sampling the full scale of meteorological regimes (e.g. clear-sky only and so sampling less of the tropical deep convective regions)?

Section 3.1 - Variability in UTH and RH in a range of datasets are shown in Figures 6-8 of Allan et al. (2022) which could be commented on. Can the shifting of wet regions more over land during La Niña and more over the ocean contribute to changes in the biases since the retrieval over land and ocean may differ subtly? Or does it relate more to the changing proportion of "dry" and "humid" regions that are sampled differently by the different instruments (particularly HIRS)? A metric for proportion of the tropics with UTH>X and UTH<X, where X could be 50% or a suitable mid-range value, would be interesting.

Fig. 2 - panels c and d do not seem to add much to a and b so could be removed since they are barely referred to

Fig. 3 - there seems to be a change in the anomaly characteristics in NCEI after 2000. Does this relate to the unusual series of La Niña events associated with slower global warming in the 2000-2012 period or are there changes in the instrument? Interestingly the anomaly characteristics of the microwave data after 2000 seem more like the pre-2000 NCEI record than the coincident post-2000 period. A similar change in anomaly characteristics seems present in Figure 8c of Allan et al. (2022) https://doi.org/10.1029/2022JD036728 for zonal means with less positive anomalies after around 2000, though it is rather a subtle change (particularly a decreasing trend in UTH 30-60^oS with positive anomalies before 2000).

Fig. 4/5 and 6/7 - a scatter of precipitation anomalies (possibly as % changes) verses UTH anomalies with some quantification of the relationship may be instructive and quite novel. It is not clear what the goal of the El Niño and La Niña comparisons are since these teleconnections are well known. If the goal is to evaluate the differences between datasets it is not so obvious from these plots. La Niña minus El Niño may be another way to present similar information in half the number of plots.

Fig. 8 suggests that CMSAF trends are inconsistent with the other datasets with increases in UTH over the whole region. Can this be quantified for the whole tropics and is this understood as the change in satellite offsets as implied. Was there no attempt to intercalibrate the satellite records by CMSAF?

Conclusions - some statements on the strengths and weaknesses of the datasets and possible issues identified would be welcome.

L324 - state that this is the CMSAF dataset which exhibits moistening trends relative to the other datasets

Will future work make comparisons with reanalyses and CMIP6 models?

Assessing the consistency of satellite-derived upper tropospheric humidity measurements by Shi et al. 

This study inter-compared the polar-orbiting satellite UTH datasets by including four participating datasets, two of which are new datasets and two of which have updated versions. Further authors have presented case studies for the El Nino period in terms of time series and spatial anomalies analysis, and the authors also try to conclude that during both El Niño and La Niña events, the values of the spatial anomalies in the infrared dataset appear smaller than those in microwave datasets and the spatial patterns of the four datasets are generally consistent over the deep tropics. This information is very informative but readers of the remote sensing community or users of these datasets might be interested to know about the quality of these datasets for their further utilization. It is suggested that include the quality of the datasets as per your results and also mention the finding of any systematic biases over different geographical regions in the manuscript. In that respect the present study has potential for publication after incorporation of the comments/suggestions as given below:

Comments:

The abstract is not informative enough.

The conclusion is not informative enough.

No discussion about insitu measurements as well as reanalysis data that provides UTH in the introduction section (e.g. Radiosonde, etc.).

The four datasets appear to be separated with two groups of similar UTH values. The values of CMSAF and FIDUCEO UTH are larger than the values of NCEI and UMIAMI UTH.

Why have authors taken the UMIAMI dataset as a reference for intercomparison? kindly add a magnitude of bias and also add some earlier analysis or findings with references.

Why did CMSAF show a negative anomaly during the year 2000 as compared to the other three datasets? Kindly explain.

Line160: adds some suitable references.

Why does the NCEI UTH trend underestimate higher magnitude over the Tibetan Plateau as compared to the other three datasets?

Line304 Over the Indian Ocean, decreased UTH centered over the equatorial central Indian Ocean is surrounded with increased UTH in all datasets.

Such types of features are not seen in the CMSAF data plot and inconsistent trends are seen in the CMSAF data plot (Fig8). kindly explain.

Reviewer’s comments on “Assessing the consistency of satellite derived upper tropospheric humidity measurements” by Shi et al.

General comments:

This article describes an assessment of four UTH datasets: one using IR measurements and the other three using microwave (MW) measurements. The main purpose is to evaluate the consistency among these UTH datasets. Focus is placed on the tropics.  Comparisons show that the four datasets are consistent in tropical-mean, interannual variability. For spatial patterns, they show broad consistency in ENSO-related action centers (e.g., Nino 4 region). Spatially, the IR-based UTH shows smaller variability than the MW-based UTH. The authors also re-examined a previous finding that was based on IR data (Shi et al. 2018), namely, upper tropospheric moistening during the El Nino events where convection is concentrated, but overall upper tropospheric drying when averaged over the whole tropics. They found that this conclusion is supported by the other three MW datasets. Finally, they analyzed the long-term changes as depicted in the four datasets by focusing on consistency among them.

Overall, I believe results from this assessment study should be useful to researchers in climate community who wish to use these datasets for diagnostic studies or model evaluations. It will contribute to the literature. As such, the paper is publishable. This kind of technical analysis fits the scope of AMT.  Therefore, I’d recommend the paper be published after minor revisions.

Specific comments:

L223-224: While UTH anomalies in the IR measurements are weaker than those in the MW in the Hovmoller diagram, the tropical-averaged UTH anomalies in the IR seem to be larger than those in the MW, as seen in Fig. 1(b). I see deeper dips in the IR data during the 1997-1998, and 2015-2016 ENSO events.

L224 – 225: “Differences in the definition and computation of UTH, the sensitivity of different sounder, and clear-sky process may all contribute to the different strengths of derived anomalies”. More details should be given to each of these causes as to how they affect the magnitude of the anomalies.  This is a technical paper. The readers will likely care about such details.

L300: Since Nino 4 region has been mentioned and used for comparison in quite a few places, it makes sense to mark it up on the maps using rectangles.

L315 Conclusion: for a technical paper, I’d recommend the main findings be organized in bullet points to facilitate reading.